Researchers find that Uber and Lyft may not be as race-blind as many hoped.
When ride-hailing services like Uber and Lyft first emerged, many hoped that, in addition to the greater convenience, users would have access to a taxi alternative that’s less prone to racial discrimination. African Americans have long struggled with racism when trying to hail a cab: Comedian Hannibal Buress jokes about it, Phoebe Robinson and Jessica Williams talk about it on the podcast 2 Dope Queens, and ESPN’s Doug Glanville has written about it in The Atlantic. In 2012, Latoya Peterson, who’s now at ESPN’s The Undefeated, wrote of Uber that ”[t]he premium car service removes the racism factor when you need a ride.” In 2014, Jenna Wortham, a staff writer for the New York Times Magazine, agreed that ride-hailing services ushered in a quality of life improvement in a Medium post. But she also raised concerns about whether the discrimination was gone or just less obvious:
It’s also not entirely clear that Uber’s system is completely foolproof. Because drivers can reject riders for any reason, you have no way of knowing whether it’s because of your rating, your name (from which race can often be inferred), or the neighborhood you’re in.
Turns out, Wortham was on the money. A new working paper in the National Bureau of Economic Research finds that for African Americans, these trips are by no means discrimination-free. A black man calling an Uber in Boston is three times as likely have his request cancelled than a white man.
The researchers found no such racial discrepancies among Lyft users in Boston—but that doesn’t necessarily mean there was no racism involved. Lyft drivers can see the names and pictures of riders before they accept, while UberX drivers cannot. “We surmise, that given that names and photos are visible to the driver prior to acceptance, any discrimination occurs prior to accepting the initial request.”
The researchers, who hail from the University of Washington, MIT, and Stanford University, conducted their controlled trials in two cities. In Seattle they got a set of diverse research assistants to take Lyft, Uber, and Flywheel rides along pre-determined, randomly assigned routes. In Boston, they asked these assistants to use “white sounding” and “distinctively black” names (as defined in a 2004 Harvard study on racial disparities in test scores) in their Lyft and Uber profiles, and then do the same. Working with a total of 1,500 resulting rides across the two cities, they analyzed the wait and travel times, cancellation rates, cost, and rating given to the rider.
In Seattle, they found that African-American riders were likely to face a 29 to 35 percent longer delay in having their requests accepted by UberX drivers. For Lyft, the “effect was too imprecise.” But again, the researchers pointed out that Lyft drivers can veto riders they don’t want to serve to begin with based on names and faces they see on their screens.
In Boston, they found that African American UberX riders overall were twice as likely to have their requests cancelled. Most of these canceled rides were in sparsely populated areas,“perhaps because drivers in those areas self select to reduce their interaction with African Americans.”
The researchers didn’t find any evidence for discrimination among Flywheel riders. They mulled over why in the paper:
There are two plausible explanations. The first is that the Flywheel service does not include photos of travelers in their profiles and so cannot present these to drivers. Therefore racial discrimination would have to be based on the names of riders. While some of the names of our African American [Research Assistants] may have provided drivers with a signal of race, we did not design the experiment to specifically test for this. Second, Flywheel works with existing taxi drivers. It is possible that the subset of taxi drivers who opt into using it are less inclined to discriminate than those who do not opt in. Perhaps taxi drivers inclined to discriminate find it easier to do so by looking at would-be passengers on the street.
Another finding seemed to indicate gender discrimination: For women riders in Boston, rides tended to be longer and more expensive, even though they had the same origin and destination as male counterparts. “The additional travel that female riders are exposed to appears to be a combination of profiteering and flirting to a captive audience,” the authors note in the paper.
In a statement to CityLab, Rachel Holt, Uber’s head of North American operations, said, “Discrimination has no place in society, and no place on Uber. We believe Uber is helping reduce transportation inequities across the board, but studies like this one are helpful in thinking about how we can do even more.” Meanwhile, Adrian Durbin, a spokesman for Lyft, said, “Because of Lyft, people in underserved areas—which taxis have historically neglected—are now able to access convenient, affordable rides.”
There’s independent evidence for the claim that ride-hailing services are indeed seen as a benefit for communities of color, and Uber has pointed to its own studies to show how it offers better service than cabs in lower-income neighborhoods in New York, L.A., and Chicago. But this research doesn’t dispute either point. What it says is that Uber and perhaps Lyft haven’t yet designed their platforms and practices to sufficiently to weed out racism—and that needs to happen if they’re really committed to serving all Americans.